Reducing VA No-Shows: Evaluation of Predictive Overbooking Applied to Colonoscopy

Patient "no-shows" are especially common in VA gastrointestinal (GI) endoscopy units, where both open-access endoscopy scheduling and patient dislike of procedures contribute to high absenteeism. In this proposal, investigators use endoscopy as a case example to evaluate a predictive overbooking model derived using patient-level predictors of absenteeism. The no-show overbooking intervention employs a logistic regression model that uses patient data to predict the odds of no-showing with 80% accuracy. These projected no-show appointments will be overbooked by clerks for patients who agree to join a "fast track" short-call line. However, patients scheduled for upper endoscopies in the "fast track" assume a small risk of service denial on the day of their overbooking in case of inaccurate predictions. If this occurs, the patient is guaranteed service in the next available position and is assured of having a shorter wait time. Patients scheduled for colonoscopies will never be turned down but may experience delays in the waiting room the day of their "fast track" appointment. By rapidly processing endoscopy patients and moving them out of traditional slots, investigators predict more scheduling slots would become available for patients awaiting colonoscopy. Investigators propose to conduct a prospective, 24-month, interrupted time series (ITS) trial in the WLAVA (West Los Angeles Veterans Administration) GI clinic endoscopy unit. During intervention periods, investigators will activate the no-show predictive overbooking strategy described above. Investigators will compare outcomes between scheduling strategies, including differences in percent utilization of capacity (primary outcome), number of Veterans served, mean patient lag time between scheduling and procedure, number of unexpected service denials ("bumps") from no-show predictive overbooking, and direct costs of care. Investigators will analyze differences using both traditional univariate and multivariate approaches, and using autoregressive integrated moving average (ARIMA) analyses to adjust for auto-correlations in ITS data.